Compared to other high-prevalence countries, Ethiopia has achieved a remarkable reduction in HIV incidence in recent years. However, the Federal Ministry of Health recognizes that inadequate prevention of mother-to- child transmission of HIV, often through suboptimal infant feeding practices, has hindered this effort. While some correlates of a key but underutilized positive feeding practice, exclusive breastfeeding (EBF), have been identified, the mediating factors that might elucidate causal relationships remain largely unknown. This study will use the Theory of Planned Behavior to examine the factors that underlie women's infant feeding decision- making and behavior, so that effective health interventions can be designed and implemented. The aims of this study are: 1) to develop and validate a culturally relevant survey questionnaire, based on the TPB, focusing on EBF among HIV-positive mothers in Addis Ababa; 2) to assess the ability of the TPB to explain EBF behavioral intention and actual behavior at time of birth and three months postpartum; and 3) to identify variables that modify the strength and stability of EBF behavioral intention. The target population is HIV-positive pregnant and postpartum women in Addis Ababa, where HIV prevalence is highest in Ethiopia. Using a purposive sampling methodology, 168 patients will be recruited from 4 hospitals in Addis Ababa. The first aim will be addressed by conducting qualitative interviews to elicit salient behavioral, normative, and control beliefs related to EBF practice among the target population. Results will be used to develop and validate a culturally relevant survey questionnaire that will allow for quantitative measurement of causal relationships between TPB constructs over time. The second and third aim will be addressed by administering this quantitative survey to a prospective cohort of HIV positive women at time of birth and 3 months postpartum. The survey will collect data on demographic and socioeconomic character, utilize Likert scales (based on formative research) to measure attitudes, social norms, perceived behavioral control, and behavioral intention, and determine actual infant feeding behavior. Data will be analyzed using Structural Equation Modeling (SEM), a technique similar to regression modeling that allows multiple independent and dependent variables to be analyzed simultaneously.